I started playing with Claude.ai and this is what I learned.
As is typical with my blog posts, this is a developing story, and I will continue to update this post as I play and learn more. For now, I am still impressed by the quality of the answers. Particularly, coding appears to yield far superior results compared to ChatGPT 3.5, ChatGPT 4, or ChatGPT 4.
Sign-up, be careful
As usual, I created a first account with dummy credentials rather than creating the account “properly”. I was so impressed with Claude.ai’s Sonnet 3.5 that I wanted to sign up. I even cancelled my OpenAI subscription to subscribe to the paid plan.
However, creating a “dummy” account first was a mistake, as of June 2024, you can’t change the mobile number associated with the account nor can you change the email address. In the antropic FAQ, they recommend that you change your email by deleting the account and making a new one.
As you need to also register a unchangable phone number that they ask you to have long term access to this becomes a problem.
Both before and after deleting the account i was unable to use the number again. Which leaves me locked out.
They use this method to prevent spam and likely also to prevent multiple sign-ups. However, I’m quite sad that this prevents me from signing up now.
Usage
Many of the prompt information I previously posted over at my ChatGPT post are still valid also for Claude.ai, as this is more general to LLMs rather than a specific product.
Prompt Library
They provide an interesting prompt gallery. I didn’t yet have the chance to use it, but I will during a quiet moment.
https://docs.anthropic.com/en/prompt-library/library
Prompt Lenght
If you keep a long chat going, it will process the context and deplete the quota and tokens faster. Maintaining a smaller context is beneficial whenever possible.
Prompt Engineering
Use XML tags to structure your prompts
Using XML tags in your prompts can enhance clarity by clearly delineating different parts of your message, ensuring a well-structured format. This method also aids in accuracy, minimizing errors that could arise from misinterpretation. Additionally, XML tags provide flexibility, allowing you to easily find, add, remove, or modify sections of your prompt without the need for extensive rewrites. Furthermore, when used in responses, XML tags improve parseability, making it simpler to extract specific information through post-processing.
To maximize the benefits of XML tagging, consider some best practices. First, consistency is key—stick to the same tag names throughout your prompts and reference those names when discussing related content, such as referring to the contract with tags. Next, employ nesting strategies for hierarchical content by using structures like . For advanced users, combining XML tags with techniques like multishot prompting and chain of thought can result in highly structured and effective prompts.
https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/use-xml-tags
You’re a financial analyst at AcmeCorp. Generate a Q2 financial report for our investors.
AcmeCorp is a B2B SaaS company. Our investors value transparency and actionable insights.
Use this data for your report:<data>{{SPREADSHEET_DATA}}</data>
<instructions>
1. Include sections: Revenue Growth, Profit Margins, Cash Flow.
2. Highlight strengths and areas for improvement.
</instructions>Make your tone concise and professional. Follow this structure:
https://docs.anthropic.com/en/docs/build-with-claude/prompt-engineering/use-xml-tags#example-generating-financial-reports (2024-10-22)
<formatting_example>{{Q1_REPORT}}</formatting_example>
Coding
I like to build some tools and utilize LLMs to code. Once you start to structure your code, it can become harder to explain to the machine about all the files when asking a question or checking for suggestions on how to modify the code.
I did create this small helper app that will combine all the files and add information about the path and filename. So I can reference the files, and the LLM will know what I mean. I still regularly use Claude AI for coding assistance as it feels superior to Copilot even though it’s not integrated into the IDE.
Feed a full codebase as with Repopack
https://github.com/yamadashy/repopack
Professional
You can use Claude.ai, also for free. However, this LLM is working so well that I’ve signed up for the professional plan.
If you are in a company, a Team tier could also be interesting, but as for me, this is personal education and fun, I only went with professional.
Unexpected capacity constraints
I’m a professional user, and even though I pay for my plan, I ran into a capacity shortage on September 2, 2024.
November 25th, 2024, and this is getting worse and worse.
The capacity issues on Claude are a continued hassle, especially considering I have a paid plan.
Claude Max
Now with the introduction of the claude.ai max plan, I’m seeing the usage limit hit, and it blocks all further queries. In the past, it seemed to switch to concise mode, but now it switches to unusable.
This is a very sad development. As a paid subscriber, I think the limit of the regular pro plan is now too low. Funny enough, this is making me a ChatGPT and DeepSeek user again.
I canceled my subscription due to the usage limits in April of 2025.
Max is coming in with a capacity 5 to 20 times higher. However, it also comes with a $100 per month plan. This is too much for me and my private use.
API
I do have access and sent a first successful request. I think as of April 2025, Claude sonnet 3.7 is still my favorite model. I’m not sure if via the API, you can trigger the longer thinking. But it would be interesting to figure out. For the moment, however, I don’t have a good use case to try Claude with.
